Re: Six Sigma and XP Estimation / Planning

At some risk of misunderstanding and misuse of the 
terminology, let me try to provide a Six Sigma perspective on 
what I think you are asking.

Six Sigma uses various processes, the most popular which is 
DMAIC (Define, Measure, Analyze, Implement, Control), which is 
used to improve an existing process. Since most of the work you 
are discussing seems to involve the development of new 
software, the more appropriate Six Sigma process would be 
DMADV (Define, Measure, Analyze, Design, Verify), which is the 
Six Sigma process for developing new products, processes, 
services and/or businesses.  

The Define and Measure step are focused on understanding the 
business environment and capturing the customer 
requirements.  Six Sigma is very focused on understanding and 
meeting or exceeding customer requirements. From that 
perspective, the extremecustomering / XP approach of including 
a representative of the customer in the development process is 
fantastic!

In the situation under discussion, it appears that the customer 
requirement that you are addressing is the development time:

===================== 
"In particular, I'm wondering about the practice of committing to a 
velocity for iteration X that's equal to the velocity achieved in 
iteration X-1.  As I said earlier, I have a feeling that could be 
perceived as "shooting low."  I wonder if 6S's statistical 
underpinnings might recommend a different approach that 
wouldn't put any more pressure on the development team re: 
unrealistic committments but, at the same time, answer this 
question plus the question about whether / how many stories 
should be estimated and in the queue in case the team 
accomplishes more than initially committed in an iteration."
======================
From a Six Sigma perspective, the issue is one of  
understanding and modeling the uncertainty - which we equate 
to variability, or the statistical term, variance.

In other words, if you knew with certainty the time it would take for 
iteration X, given the known velocity achieved in iteration X-1, then 
there would be no problem with projecting the development time.

So, we would approach this as the need to develop a statistical 
model (also called a "stochastic model") for the development 
time. 

There are several ways to develop such a model, but let me start  
with a fairly simple model that derives from the description you 
give.

Let's imagine the sequence of steps is the iterative process you 
describe:

Iteration 1 ---> Iteration 2 --....-->Iteration X-1 ---> Iteration X 
--->Iteration X+1.....--> End

Now, each iteration occurs at some velocity and consumes 
some amount of time. 

One statistical modeling approach would be to assume 
independence for each of these iterations. You've indicated, and 
I would also suppose, that these iterations are not independent - 
they are in fact autocorrelated (correlated with themselves: if you 
know the velocity of iteration X-1, you might have some indication 
of  what the velocity of iteration X might be). This autocorrelation 
is part of what you want to understand. 

To be honest, I'm not at all clear about the nature of the 
autocorrelation - if I speak in terms of time rather than velocity, 
then I can argue that the autocorrelation between the time for 
iteration X-1 and iteration X would be positively correlated:if it 
takes a long time to perform iteration X-1, it will probably take a 
long time to perform iteration X.

However, I can also argue a negative correlation in terms of time: 
if I spend a lot of  time on iteration X-1, doing a very thorough job, 
then it might take less time to perform iteration X.

So, just to start out, let's make an initial assumption that the 
iteration times are uncorrelated.  

With this assumption,we would estimate the distributions of 
times for each iteration, then use those distributions to project 
the distribution of the total development time.

Now, each of these iteration times would not be expected to 
follow normal distributions. There is a minimum time that you 
would need to spend on each iteration - the minimum time could 
be zero (by contrast, a normal distribution would allow for the 
possibility of negative development times).  There would be an 
expected time for each iteration, but there would be no maximum 
time - each iteration could theoretically take forever, although that 
might make someone a little bit unhappy.  However, there would 
be an estimate of the standard deviation - the uncertainty - of the 
distribution of time for the iteration; the square of that standard 
deviation would be the variance.

This type of distribution is called a Gamma Distribution. You 
would obtain an estimate of the gamma distribution of time for 
each iteration, then obtain a new gamma distribution for the 
project - it would have a minimum time, equal to the sum of the 
minimum times for each iteration, a most likely  time, equal to 
the sum of the most likely times for each iteration, and a 
variance...equal to the sums of the variances for each iteration.

Now, there is a way to take this approach, which assumed 
independence, and adapt it to a situation where there are 
correlations.  If appropriate, I'll go into that later.

There are also numerical simulation approaches, using discrete 
event simulation tools to simulate running through the same 
sequence of  iterations thousands of times, and then seeing a 
histogram that would be an estimate of the gamma distribution 
of the total development time from that sequence of iterations.
This is a variation of a statistical approach called Monte Carlo 
simulation.

This message is getting long, so I'll try to continue it in a later 
message.

Best regards,
Eric Maass
Chairman, Six Sigma Steering Committee, Motorola SPS

--- In [email protected], "Bill Walton" 
<[email protected]> wrote:
> We're joined now by a real Six Sigma guru.  His name is Eric 
Maass.  Eric is Director of Technology Strategy for Motorola's 
Wireless and Broadband Systems Group.  At the same time, he 
is Chairman of the Six Sigma Black Belts Steering Committee for 
Motorola's Semiconductor Products Sector.   He has a web page 
at http://www.geocities.com/ecmaass/ for anyone who wants to 
check out his CV.  In a conversation with another 6S guru who 
hasn't had time to join us I was told "I think you hit the nail on the 
head. The future is really in a balanced mix of XP and Six Sigma."  
So I have some high hopes that, in talking with Eric, we may find 
that XP has an ally in 6S that could bridge some communication 
gaps with organizations that are doing 6S, or at least think highly 
of it, but who don't know as much about XP.
> 
> I posed a question earlier about whether / to what extent XP 
would find support for it's estimating / planning practices in a Six 
Sigma environment.  In particular, I'm wondering about the 
practice of committing to a velocity for iteration X that's equal to 
the velocity achieved in iteration X-1.  As I said earlier, I have a 
feeling that could be perceived as "shooting low."  I wonder if 
6S's statistical underpinnings might recommend a different 
approach that wouldn't put any more pressure on the 
development team re: unrealistic committments but, at the same 
time, answer this question plus the question about whether / 
how many stories should be estimated and in the queue in case 
the team accomplishes more than initially committed in an 
iteration.  I also tend to think XP's practice of picking a "best 
guess" at an initial velocity and then adjusting based on actual 
results (as opposed to trying to develop a long term timeline that 
will be difficult to adjust, etc.) may find support in 6S in the 
context of establishing a baseline.
> 
> Eric is new to XP and has expressed a need to come up to 
speed on the practices and the terminology.  I thought it would 
be best, since practices are evolving, since there seems to be 
some variation in practice, and since I'm still researching XP as 
opposed to actually doing it, to ask you guys to explain the 
estimating / planning practices you use.  Or, if you like, I could 
explain it as I understand it from my readings and talking to you 
here and let you correct me as necessary.
> 
> Best regards,
> Bill



------------------------ Yahoo! Groups Sponsor ---------------------~-->
Get 128 Bit SSL Encryption!
http://us.click.yahoo.com/CBxunD/vN2EAA/xGHJAA/NhFolB/TM
---------------------------------------------------------------------~->

To unsubscribe from this group, send an email to:
extremecustomering-unsubscribe-hHKSG33TihhbjbujkaE4pw@public.gmane.org

 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/